About this question

False Sharing Benchmark

Medium · concurrency · Quant Developer interview question · concurrency, performance, memory-layout, atomic

False sharing degrades performance in low-latency trading systems when multiple threads modify independent variables residing on the same CPU cache line, triggering excessive cache coherence traffic. Mitigating this requires explicit memory alignment or padding to ensure frequently updated variables are separated by the architecture's cache line size, typically 64 bytes. Task Implement the thread-safe class AlignedCounters to manage two independent int64_t atomic counters, ensuring they are sep